Hello Pix4D team,
I’m excited by the new point cloud classification update and am currently in the midst of processing this as a classified .las file.
My question is therefore if these classification attributes will be part of LAS file, so as to visualise these using other point cloud environments; i.e. plas.io, pointbox.xyz or indeed any other platform underpinned by the LAS 1.4(?) standard and similar libraries?
Thanks in advance for your time.
Paul
Having processed and used the point cloud classification; I can see that exporting the resulting .las file from Pix4D does not seem to save the classification attributes.
Am I doing something wrong here?
You can manually export the point cloud in the rayCloud editor and simply select each classification separately. Might not be the intended way to do it but new features rarely work 100% so this work-around might do the trick for you.
Agreed with Adam, here is how I would proceed:
Then this pop-up should appear:
Where you can select which point groups you want to export together in the same file.
Many thanks to both of you for the responses. Appreciate that this is still a new feature however, in the example above, are you both then recommending that I export each point group (ground, road surface, etc) individually as a means to export these classifications? Please confirm thanks :)
If this is so, I guess the only remaining question is whether Pix4D intends to align these classification codes as defined in the LAS 1.4 standard in the near future?
Please let me know thanks.
All the best,
Paul
Yes Paul, export each group into individual files.
@Paul: For the moment we use the LAS 1.3 standard, normally it should be compatible with the LAS 1.4.
In the future the plan is to pass to the LAS 1.4 standard for the classified point clouds and the point cloud exports in general. However, there is no clear estimation on when this would be for the moment, so it is not for the near future.
I am also excited by the automatic point cloud classification development. I read the white paperassociated with this great evolution: Incorporating colour and not just geometry into training certainly opens new frontiers.
Many of my projects are in lightly to densely wooded savanna, so being able to group “high vegetation” points together would greatly facilitate the DTM estimation.
Here follows screenshots of my first automatic point classification before editing anything:(Pix4D Mapper Pro Ver. 4.0.21, Fixed-wing, piloted aircraft, Canon 5d mk IV, Sigma 35mm f/1.4, 75 knots, 1 sec trigger intervals, 95m wide tracks, 850ft above ground average, resulting in GSD 4cm/pixel.)
Image 1: All point groups in the flat area of the project:
Image 2: High Vegetation point group removed:
Image 3: Only the High Vegetation group
In another part of the project there are steep and overhanging cliffs that get grouped into “High Vegetation”.
Image 4: All point groups in hilly part of project:
Image 5: “High Vegetation” point group removed from hilly part of project:
Image 6: “High vegetation” point group only.
So, in the flat terrain the classification does a remarkably good job of grouping trees into “high vegetation”.
In the very hilly terrain, however, the high vegetation group incorporates a lot of non-vegetation points (boulders, cliff faces). I imagine the brownish cliff faces with green caps are understandably confused with tree trunks?
Whatever the reason for the aggressive pro-vegetation classification, I’m sure that further training using different kinds of terrain will gradually sort this out.
A note: Early in our rainy season, the trees green up before the grass does. Grouping points purely on colour would be a simple and useful facility. This is a special case scenario but there might be other uses.
Congratulations with a fine evolution. Looking forward to its refinement.
@Wynand Uys: Thanks for illustrating your comment with these screenshots. I also think that the algorithm takes those cliff faces and boulders for trees, probably due to the color and shape that is similar.
We always try to improve our algorithms. If you’d like that the algorithm gets improved for your use case, you can do the following:
Our developers will have a look and will feed that to the algorithm. 
Thanks for your feedback!
Thanks Pierangelo
It wasn’t a big deal to correct the classification manually.
Most of my other work is in relatively flat terrain, and the algorithm does a fine job.
I’d be happy to send the dataset.
Um… how do I go about that?
I guess they’ll need the images? 990 images of 30MP = 35GB disk space slightly compressed. I have a slow connection. It might take a day or two to upload.
Sorry for the delay, I missed your reply and stumbled on it by chance. Yes, if you could share the images as well it would help. You could use Google Drive, Dropbox or a similar file sharing service and send the link to your folder in the email above. Thank you!
I was wondered…Is the DTM based on point cloud classification which I selected?
For example, what if I just want the “Road surface” & “Ground” of point cloud group to produce the DTM. Is there any way to achieve it?
I had tried to use the point cloud classification, but the final DTM seem still included the “High vegetation group”.
Normally, the DTM (Digital Terrain Model) is generated based on the Ground and Road Surface point groups by default. Maybe what you saw is the DSM (Digital Surface Model). You can read more about how to create the DTM here: https://support.pix4d.com/hc/en-us/articles/202560579
So it is. I got it.
Thanks.
I met another issue about point cloud group vs DTM.
My project includes the riverbank. There were a few weeds on the top of the riverbank. Most the top of riverbank classified as the groups of “Groud” or “Road Surface”, but some areas of the top of riverbank were classified as “Human Make Object”.
I tried to edit the wrong point cloud into “Groud” group, and re-processed Step 2 & Step 3. But…the result of DTM still didn’t include the wrong group of “Human Make Object”.
How do I correct this error?
If you run step 2 again, the classification will be run from scratch, so you will loose the edits you have added.
I would recommend to follow the steps from this article and then run step 3: How to edit the point cloud in the rayCloud.
Yes, it works. Thanks.
No wonder I always got the result which is not I expected.
Ps: After I edited the classification,and I already produced DSM. Could I just run the “Process / Generate DTM” & “Generate Contours lines(DTM),” instead to run the whole Step 3?
Because in our Workflow, we just want the DTM’s contours lines. We don’t need DSM, Orthomasic, Index…etc.
Hi Lung,
The contour lines derive from the DSM / DTM. In other words, the DSM / DTM needs to be regenerated to account for the changes made in the point cloud. This means you have to re-process step 3.
But…This Link said, “This process takes as input the merged Raster DSM (Digital Surface Model), computes a classification mask and generates the Raster DTM (Digital Terrain Model).”
If my DSM didn’t change any point cloud, I just re-edit the classification of the point cloud. That is, I changed the “classification mask”. So, I should not need re-process Step 3, is that right?
And today, I made a little experiment about this issue.
After I moved some point cloud which is in the "Ground " and “Road Surface” to another point group, and we did the shortcut of running the “Process / Generate DTM” & "Generate Contours lines(DTM). Pix4Dmapper indeed change the result of DTM’s contours lines.
It seems no need to run the whole Step 3. Would you classify this “Shortcut”? Because our project could include thousands of photos to process. We wish this shortcut could save the time of processing Step 3.
FYI, my tone may be rude to you, that is…my English is very poor. I did not do it on purpose
Hi Lung,
Sorry my mistake. You do not have to re-run the entire step 3 but you still need to re-generate the DSM / DTM. After editing the point cloud, you can take advantage of the Process menu.